►
From YouTube: Ceph Orchestrator Meeting 2023-03-14
Description
Join us weekly for the Ceph Orchestrator meeting: https://ceph.io/en/community/meetups
Ceph website: https://ceph.io
Ceph blog: https://ceph.io/en/news/blog/
Contribute to Ceph: https://ceph.io/en/developers/contribute
What is Ceph: https://ceph.io/en/discover/
A
C
A
So
yeah,
so
there
was
one
thing:
I
thought
it
could
bring
up,
which
is
I
opened.
This
book
request
yesterday.
A
I'm,
sending
the
OSD
Flags
very
upgrade,
and
so
the
thing
I
would
stop,
maybe
would
be
worked.
A
little
bit
of
discussion
is.
A
How
we
handle
sort
of
what
flags
are
set
and
what
we
do
by
default,
versus
what
they
have
sort
of
asked
for
we'll
say
the
way
that
I
handled
it
currently
in
that
pull
request
is
I
made
the
flags
that
you
want
us
to
set
the
config
option,
it's
like
a
comma
separated
list
and
then
a
system
by
default
you
do
the
upgrade
and
then
there's
a
flag.
A
A
If
you
want
something
and
then
you
can
also
just
have
it
not
do
anything
of
any
Flags,
the
reason
obviously
was
going
to
have
it
I'll
do
it
by
default
and
then,
if
they
pass
the
flag
and
then
it
would
add
them
in
the
tracker
actually
simply
asked
for
it
to
be
by
default.
That's
from
somebody
who
should
know
a
little
bit
about
this
stuff,
so
yeah.
What
about
that?
A
So
I
guess
I'll!
Just
sort
of
open
the
floor:
if
anyone
has
any
opinions
on
sort
of
good
ways
to
specify
the
flags
or
what
should
we
do
by
default?
First,
what
they
should
have
to
ask
for,
or
something
like
that
as
I
said
I,
have
it
sort
of
the
way
I
have
it
in?
There
is
sort
of
what
I
came
out
with
as
I
thought.
A
It
made
sense,
which
is
a
better
idea,
especially
for
specifying
which
flag,
because
the
common
separated
list
as
a
config
option
is
a
bit
awkward
but
I
wasn't
sure
exactly
where
to
put
it
either.
I
didn't
want
to
have
a
different
flag
for
the
upgrade
command
for
like
every
single
OSD
plug
you
want
set
no.
A
All
the
osds
so
basically
like
say,
for
example,
the
no-out
flag.
It
just
means
that
none
of
those
these
will
get
marked
out
during
the
upgrade
yeah,
so
I
mean
I,
didn't
have
to
come
up
as
much
Flags
would
be
a
good
default
either
I
took
them
directly
from
the
tracker,
which
is
I,
think
looks
like
it
was
taken
from
stephensible.
A
That's
right
where
I'm
at
with
it
is
I,
don't
know
it
just
felt
like
a
bit
of
a
strange
way
of
doing
it,
but
I
wanted
it
to
be
customizable.
I
didn't
want
to
like
force
a
certain
set
of
flags.
I
didn't
really
have
a
better
way
of
doing
it.
They
said
other
than
having
like
a
separate
flag
for
each
OST
flag
in
that
great
command,
but
I
didn't
like
that
either.
But
I
thought
we'd
just
have
way
too
many
flags
in
that
command.
D
B
Does
this
make
sense
to
pass
some
configuration
file
to
that
rate
back.
A
File
for
upgrade
I,
don't
know,
I
mean
it
feels
like
kind
of
heavy
for
that
quick
man
because
never
used
those
before
I
guess
we
have
to
compare
it
to
I
guess
what
I
have
in
there
right
now,
which
again
is
the
config
option
like
what
is
more
or
less
annoying
I,
don't
know,
I
feel
like
making
a
spec
file
for
the
upgrade
might
be
actually
harder
than
setting
the
config
option,
especially
if
you
were
okay
with
the
defaults
we
have
set
up
there
and
you'd
have
to
sort
of
still
make
a
spec
file.
A
D
D
D
Yeah,
well,
that's
what
I
was
assuming
I
I
was
assuming
it
wasn't,
something
people
would
be
tweaking
a
lot.
D
A
C
A
A
So
maybe,
if
you
have
all
of
those
things,
maybe
something
we
would
think
about.
There's
enough
things
that
it's
worth
having
a
spec
yeah
I'm,
just
gonna
hit
like
10
of
them
on
just
because
it's
not
what
we've
done
before
and
we'd
have
to
then
like
explain
how
to
set
up
this
new.
D
D
A
A
A
A
D
A
A
So
that's
why
I
was
originally
going
to
make
this
off
by
default,
nearly
had
to
pass
the
flag
to
do
it
because
then
like,
if
you
pass
the
flag,
you
must
oh
there's
something
going
on,
but
so
the
tracker
was
recommending
to
do
it
by
default.
I
think
that
is
what
stephensible
is
doing.
I
think
it
was
setting
these
five.
A
It's
just
that
we
probably
should
have
had
this,
maybe
even
earlier,
and
we
didn't,
but
it's
just
I,
don't
know
the
reason
why
it
turns
out.
It's
been
so
long
now
that
it's
worked
like
it's
worked,
that
people
might
be
just
sort
of
surprised
that
in
a
different
now.
A
So
we
say
like
this
is
a
major
release
thing
because
it
also
won't
trigger,
and
you
upgrade
to
a
version
that
has
this
changed.
It
has
to
be
like
from
so
we
waited
until
they
like
the
brief
release
for
this
to
actually
hit
an
actual
release.
Then
at
least
it
would
be
like
a
new
reef
feature.
It's
not
like.
Suddenly
in
your
Quincy
Point
upgrade
it.
C
A
And
that's
maybe
somewhere,
where
you'd
be
more
willing
to
accept
like
all
right.
This
is
a
bit
different.
I
know
they
have
like
they're
gonna,
have
like
release,
notes
and
stuff
a
bit
more
attention
to
on
the
major
releases.
D
A
D
A
I
could
do
that?
That's
that's
a
that's
true,
a
probably
a
good
idea.
I,
don't
know
what's
easier
and
that
change
in
the
backboard
would
be
very
simple.
It
would
always
be
changing
one
string,
yeah
so
technically
wise.
That
would
be
very
easy
and
I
guess.
I
would
move
the
mechanism
there,
but
then
it
would
have
no
actual
change.
D
A
Yeah
I
might
have
to
go
for
that
I
kind
of
like
that
idea,
because
if
we
don't
surprise
anyone
with
anything
because
it
won't
actually
do
anything
unless
you
know
about
it-
and
you
said
something
on
the
minor
releases
and
then
on.
The
major
release
we
can
have
like
here
is
pronounce
that
these
flags
and
these
like
release,
notes
and
again
I.
Think
because
it's
a
major
version.
People
are
more
willing
to
accept
things
changing.
That's
where
you
expect
there
to
be
more
significant
changes.
A
That's
fine
because
that's
what
we're
also
doing
with
like
say,
like
our
secure
monitoring
stack
stuff
that
all
relies
on
like
moving
like
retia
surface
Discovery,
from
like
your
old
style
and
like
that's.
Why
we
didn't
want
to
backboard
that
we
thought
it'd
be
too
surprising
for
our
minor
release
to
suddenly
change
that
stuff.
A
But
this
is
what
you're
talking
about
actually
happens
to
work,
for
you
can
have
the
feature
be
there,
but
not
do
anything
by
default.
I
think
actually
is
probably
the
best
option.
A
Yes,
well,
what
sort
of
tries
to
do
is
it
has
a
set
of
flags
and
things
kind
of
sets
and
it
makes
sure
they're
set
before
it
starts
upgrading
demons
and
then
once
the
upgrades
as
the
upgrade
completes,
it
tries
to
unset
all
those
flags.
So
we
do
have
the
one
problem
that
if
you
had
those
flag
sets
before
yeah,
it
will
unset
them
exactly.
C
A
Okay,
like
oh
we're,
messing
with
the
flags
a
little
bit
here,
whereas
in
like,
is
that
in
a
point
release
you're,
just
upping
a
point
release
you
don't
expect
there
to
be
anything
super
significant,
and
so
they
would
I
don't
know.
Maybe
that
would
be
a
bit
bigger
deal
but
I'm
not
sure
how
much
of
a
use
case
that
is,
or
people
just
like
go
out
set
all
the
time.
Maybe
there's
something
wrong
with
their
osds
or
something.
A
I,
don't
think
I'm
saying
the
flags
went
out
with
that,
because
that
would
be
a
promise.
You
would
actually
want
not
off
at
that
point,
if
you
want
those
to
be
marked
out.
B
A
It's
very
possible
I
know
so
one
of
the
things
that
it
does
when
it's
going
to
the
loop
to
set
the
flags,
it
doesn't
just
try
to
set
them
all,
because
it
is
item
potent.
So
it
could
do
that,
but
it
doesn't
have
to
do
that.
It.
It
checks
if
they're
there
first.
A
So
we
could
do
that
like
initially
and
then
save
that
as
part
of
our
upgrade
State
like
what
the
initial
date
was
and
then
sort
of
this
that
against
what
we
need
to
set
yeah,
that's
probably
possible
I
think
it's
I
think
it's
doable.
We
think
that
it's
like
a
it's
too
risky
to
unset
people's
Flags.
A
Okay,
I
mean
I'm.
Okay,
with
that
I
think
I
could
I
could
take
a
look
at
that
and
see
how
doable
it
is.
I
think
I
think
it
would
actually
be
pretty
easy,
but
just
be
one
extra
thing:
we
have
to
store
in
the
upgrade
state
and
then
at
the
end
we
just
basically
not
do
unsets.
If
we
saw
those
flags
were
in
that,
like
initial
list
that
we
picked
up.
A
All
right,
I
think
that's
all
I
had
about
this
floor
class
I
just
want
to
ask
a
couple
questions
about
it,
because
it
was
doing
some
weird
things
by
default
now,
not
weird
things
but
doing
things
that
are
different
from
what
I
was
doing
before,
but
I
kind
of,
like
both
of
these
things
we
talked
about.
Maybe
the
restore
thing
is
probably
good
and
then
the
backboarding
that
the
defaults
is
probably
nice
as
well
bring
the
feature
back
without
actually
changing
any
Behavior
by
default.
A
C
Okay
hi,
so
my
name
is
Gotham
and
I
I've
been
in
this
meeting
before,
but
I'm,
not
a
frequent
attendee,
so
I
I
work
on
the
on
openstack
manila
and
we
had
some
open
trackers,
and
this
was
us
testing
with
the
Ingress
and
the
ceph
NFS
service
about
I
mean
a
few
months
ago
or
something
like
about
last
year
more
like,
and
so
we
we
realized
that
the
the
the
English
service
was
was
actually
preventing
client
restrictions
to
be
applied
on
access
rules
so
on
on
exports
that
were
generated
by
the
Steph
NFS
service,
and
so
we
were
working
with
the
NFS
Ganesha.
C
You
know,
developers
took
to
kind
of
come
up
with
a
a
a
solution
to
this
and
the
they've
gone
ahead
and
implemented
the
proxy
proxy
protocol,
so
that
that
would
allow
the
NFS
connection
to
actually
see
the
client
address.
That's
the
the
real
client
address
instead
of
the
Ingress
address
that
it
currently
sees,
and
that
would
allow
the
client
restrictions
to
work
again.
C
So
there's
a
little
bit
of
background
of
the
things
that
we
were
looking
at
last
year,
and
so
now
we
just
we.
We
have
some
NFS
connection,
builds
to
play
with
and
I
think
it's
time
to
go
back
to
Safe,
Ingress
and
and
do
some
experiments
so
come
here.
Give
you
a
heads
up
Adam
about
a
couple
of
trackers.
C
So
the
first
one
there
is
the
NFS
Ganesha
issue
that
tracks
the
proxy
protocol
implementation.
It's
not
closed
yet,
and
that's
because
we're
we're
still
testing
this
stuff
and
I
mean
there's
a
PR.
That's
linked
out
of
there
and
there
is
the
the
the
seph
Adam
tracker.
C
C
And
the
the
config
option
is
proxy
underscore
hosts
and
it
it's
basically
the
in
the
Ingress
IPS
or
I
mean
if
there
are
multiple
the
IP
addresses
of
the
Ingress
service.
C
So
that
Ganesha
can
know
that
the
request
is
coming
from
NHA
proxy
and
and
can
look
at
the
in
the
header
for
the
actual
client
address
and
in
the
Ingress
service
itself.
The
HR
proxy
config
would
need
to
have
another
value,
which
is
to
send
the
proxy
protocol
header
and
that's
detailed
as
well.
So
this
this
is
I,
think
the
changes
that
we'd
require
to
kind
of
integrate
the
the
solution
completely.
A
I
want
to
say
right
so
there's
two
changes.
We
need
there's.
C
A
A
Nfl
is
like
something
called
like
proxy
hoax.
What
was
the
thing
called
that
we
had
to
in
this
tracker.
C
On
the
tracker
that
that
you
have,
let
me.
A
Four
and
then
in
the
assumes
the
proxy
config
yeah
you
need
something
as
well
is.
C
That
also
one
of
those
sorry
I
I
cut
you
off
there.
A
Oh
so
there's
the
is
the
stuff
for
the
AJ
proxy
config.
Also
one
of
these
tracker
links,
yeah.
C
It's
in
the
in
the
same
tracker
did
you
want
me
to
create
a
separate
one?
No.
A
C
Yep,
so
it
I
can
give
you
an
IP,
so
yeah
a
different
link
there.
So
it's
a
single
line
called
send
proxy
V2
and
I'm
just
trying
to
find
that.
C
C
Yes
and
that's
kind
of
what
we
were
trying
to
set
up
for
right
now,
like
we
actually
reached
out
to
somebody
the
internally
in
the
chef
team
and
and
I
checked.
If
we
can
get
a
build
with
the
NFS
connection
packages
that
showed
up
on
their
their
currently
pushed
out
to
CBS.
A
A
A
Okay
but
yeah,
this
is
not
super
complicated
because
they
said
because
we
use
the
most
recent
one
we
could
typically
I
should
be
able
to
just
make
a
four
of
us
against
Maine.
That
makes
these
sort
of
configuration
changes
and
then
we
can
sort
of
see
how
it
works.
A
Okay
and
then
this
other
one
about
the
port
stuff,
is
that
part
of
this
is
a
separated.
C
Yeah
right,
yes,
so
we've!
This
is
another
thing
that
we
may
run
into
and
we're
not
entirely
sure.
So
it
was
a
separate
issue
and
we've
discussed
this
here
in
the
past.
C
C
As
we
can
see,
I
I
see
there
are
options
to
do
that
when,
when
setting
up
the
NFS
service,
but
I
I
mean
the
NFS
Ganesha
team
thought
that
the
RPC
bind
would
actually
be
advertising
the
the
actual
NFS
connection
Port
and
might
prevent
you
know
for
prevent
us
from
mounting
on
that
2049
Port.
So.
A
A
A
Always
a
conflict,
yeah,
I,
think
yeah,
so
I
know
so
I
think
we
have
someone
else
who
was
going
to
look
at
this
who's
trying
to
get
into
more
stuff.
Idm
development
was
checking
this
one,
so
this
one
we
will
figure
out,
but
I
just
want
to
confirm
that
this
is
not
going
to
block
any
of
the
or
shouldn't
hopefully
block
any
of
like
this
proxy
stuff.
This
is
just
like
a
separate
issue.
It.
A
Yeah,
okay,
I
think
we
have
somebody
who's
looking
into
that,
but
I
might
try
to
do
this
process
configuration
stuff
first
and
then
we'll
get
back
to
the
the
port
issue
after
so
I
think
this
is
the
bigger
thing.
Actually
is
this
because
you
can
just
configure
around
the
the
port
stuff
really
want
to.
A
A
All
right
now,
so
it's
not
too
bad.
It
sounds
like
you
already
have
all
the
links
and
stuff
that's
what
we
have
to
do
so.
I'll.
Take
a
look
at
that.
If
I
can
make
a
test
request
to
change
those
things
and
see
how
it
goes.
A
Right,
anyone
have
any
else
on
that
or
any
other
topics.
A
In
that
case,
we'll
we'll
call
it
here,
get
a
half
an
hour
back,
yeah
and
and
thanks
for
coming
I'll
see
you
guys
next
week.